2015 19th International Conference on System Theory, Control and Computing (ICSTCC), October 14-16, Cheile Gradistei, Romania

HVAC Control Methods - A review

Filip Belic Zeljko Hocenski, Drazen Sliskovic
Design and Construction Department of Computer Engineering and Automation
TEO-Belisce Faculty of Electrical Engineering
Belisce, Croatia Osijek, Croatia
filip.belic@teo-belisce.hr zeljko.hocenski@etfos.hr, drazen.sliskovic@etfos.hr

Abstract—Making buildings more energy efficient while
keeping thermal comfort has been an important topic in last
decades, including improving efficiency of Heating, Ventilation
and Air Conditioning (HVAC) systems. One approach to this
goal is to use better control methods. This paper presents a
review of methods which have been researched in scientific
papers in last couple of years. Most papers focus on one of the
following trends: improving "classical" control methods; use of
predictable control based on models; use of intelligent control
methods.
Keywords—HVAC systems, control, review, Classic Control,
Predictive Control, Intelligent Control
noninvasive and cheaper approach. At the same time, these
control methods must provide existing or even upgraded feel of
comport to occupants, which is a primary goal of HVAC
system.
This article presents an overview of some recent researches
that deal with improved control methods in HVAC systems.
Second part of paper presents some general topics dealing with
control strategies in HVAC systems. Reviewed papers are
presented in this part, divided to groups by approach. Third
part will try to find some common characteristics and trends in
presented articles. Final part will give conclusion and propose
possible further steps.

I. INTRODUCTION II. CONTROL STRATEGIES FOR HVAC SYSTEMS

Energy efficiency is a popular topic in last decades, both in As already mentioned, HVAC originates in Heating,
scientific and popular literature. Data [1], [2] shows that Ventilation and Air Conditioning. Nowadays, HVAC system is
developed countries use up to 40% of energy on buildings, and considered a part of system called air-conditioning system,
that almost half of it is used for heating, ventilation and air which takes care of the total control of temperature, moisture in
conditioning (HVAC systems, henceforward). This means that the air (humidity), supply of outside air for ventilation,
up to 20% of energy in USA today is used for HVAC systems, filtration of airborne particles, and air movement in the
shown in Table 1. And if we consider that total energy occupied space [4]. Nevertheless, HVAC is still used in most
consumption is 28×1012 kWh (data for USA in 2010), this literature. HVAC comprises following processes:
proves quite a large amount.
 Heating – addition of thermal energy (heat) to air with
purpose of maintaining zone's temperature
TABLE I. ENERGY USE IN USA, PER CONSUMERS OVER YEARS
Energy used
 Cooling - removal of thermal energy from air with
Consumer / Year 2006 2008 2011 purpose of maintaining zone's temperature
Buildings (% of total) 39% 40% 41%
 Humidifying – addition of moisture to air with purpose
Residential (% of buildings) 54% 54% 54%
Commercial (% of buildings) 46% 46% 46%
of maintaining zone's humidity
Space heating (% of buildings) 20% 21% 37%  Dehumidifying – removal of moisture from air with
Lighting (% of buildings) 18% 13% 9%
purpose of maintaining zone's humidity
Space cooling (% of buildings) 13% 13% 10%
Water heating (% of buildings) 10% 9% 12%  Ventilating – replacing of interior air with fresh air
Other (% of buildings) 39% 44% 32% from outside with purpose of maintaining zone's air
HVAC (% of buildings) 33% 34% 47% quality
HVAC (% of total) 13% 14% 19%
 Cleaning – removal of dust and other particles with
purpose of maintaining zone's air quality
There are several approaches to reducing this energy and
making buildings more energy efficient: use of modern  Air movement – circulation and mixing of air with
materials with better thermal characteristics; applying design purpose of maintaining zone's temperature, humidity
methods that consider energy efficiency during design phase; and air quality
replacing equipment with more efficient one; use of better
control methods; etc. [3] Some of these methods can be applied This article focuses only on first four or five processes of
only before and during construction; some of them are HVAC systems (heating, cooling, humidifying, dehumidifying;
expensive to apply. Upgrading HVAC system to be more sometimes ventilation).
energy efficient by use of modern control techniques can be

978-1-4799-8481-7/15/$31.00 ©2015 IEEE 679

A. Approaches to Control of HVAC Systems They explain how to use decoupled approach to realize
For analysis of control methods for HVAC systems, it is decentralized control system. A feedforward controller in
first necessary to make a distinction by approaches. combination with PID controller is selected as showcase,
demonstrated on refrigeration system. Results show
In paper [5] author shows development of control methods improvement of RMS (Root Mean Square) error of control
in industry (not just in HVAC) through history. Control system and more stable operation of components.
strategies are developed in four stages, which can be seen in
Table II. First stage is labeled Conventional PID Control: In paper [7] authors propose and discuss an improvement of
second stage (Advanced Control I) is marked by different existing optimal control of heat-exchangers in HVAC systems.
versions of Multiple Input Multiple Output Control; stage three Authors shortly discuss model of two heat-exchangers (air-to-
(Advanced Control II) comprises different types of industrial air and water-to-air) and give a detailed control strategy:
implementations; fourth stage (Advanced Control III) is signed disturbance rejections; maximum exploitation of air-to-air heat-
by different hybrid, expert and soft-computing techniques. exchanger; keeping difference between supply water flow and
Second part of article deals with trend of implementation of tertiary water flow of water-to-air heat-exchanger on minimum.
various ICT (Information and Communication Technology) They discuss classical and optimal approach based on two
techniques in process control. independent controllers for heat-exchangers, conditions and
constraints for control. Next they propose a simplified optimal
control scheme, which solves problem of bypass flow, resulting
TABLE II. DEVELOPMENT OF CONTROL STRATEGIES OVER TIME in 82% reduction of energy consumption of tertiary pump.
A B C D Simple application of PID control is given in paper [8],
Conventional Advanced Control Advanced Advanced
PID Control I Control II Control III where authors explain design of handheld instrument for
Robust Hybrid automatic tuning of PID controllers that can be used in HVAC.
Manual Gain Scheduling
Control Method
Control Predictive They give theoretical background for tuning PID controllers
Methods Control and explain how it can be done based on process data. Paper
Adapting and Optimal [9] explain use of PID controllers for control of HVAC system
Feedback Fuzzy
Self-tuning Control
Control
Control Methods
Control of large commercial building. PID control is combined with
Multivariable expert knowledge to provide better results: simulating human
Control Methods Model Neural divisional control (by providing different control scenarios);
Cascade
Control
(State Space and Predictive Network on-line tuning of PID parameters (depending on accumulative
Transfer Function Control Control error); feed-forward control (depending on outside conditions);
Models) and lag compensation of particular HVAC sub-systems.
Multivariable
Control Methods Discrete Paper [10] present a detailed explanation of classical
Feedforward Decentralized
(Decoupling and Events control strategy based on PID controllers. Authors first
Control Control
Decentralized Control
Control) describe building and HVAC system and use first-principles
Nonlinear approach to build a thermal model which comprises all relevant
Ratio Control
Pole Placement Algebraic Hybrid Soft thermal processes. They use Laplace transformations to get
Methods Control Computing transfer functions. These equations are used to model PID
Control controllers for humidifier and heating coil for heating mode,
Combined Robust QFT Expert
Control Nonlinear Control Control Control
and for dehumidifier and cooling coil for cooling mode. PID
Structures Methods Methods controllers are tuned with Ziegler-Nichols method.

To take a similar approach, this paper will divide reviewed
articles into four categories: classic approach (Feedforward and
Feedback Control, On-Off, PID, etc.); predictive control
methods (in most cases, Model Predictive Control); intelligent
control techniques (fuzzy logic in various combinations); and
other approaches and general topics.
B. Classic Approach
Also called conventional control strategies, this approach is
marked with use of methods with many decades of practical
implementation, especially feedback control and PID
(Proportional-Integral-Derivative) controllers.
Paper [6] is dealing with design of control systems. Even
though authors admit that MIMO (Multiple Input Multiple
Output) system provide better theoretical results, in practice it
can be more convenient to have several SISO (Single Input
Single Output) systems, especially when tuning is considered.
C. Predictive Control
Predictive control techniques rely on predicting the
dynamic behavior of system in future and adjusting response of
controller accordingly. Usually this means some variation of
Model Predictive Control (MPC), where prediction is
performed based on explicit model of building. These kinds of
methods can achieve very good results, but often have
problems with complex implementation. Fig. 1. shows relation
between important values of MPC: setpoints, predictions and
inputs that are adjusted based on these predictions.
Book [11] gives very extensive overview of Model
Predictive Control, with focus on industry. MPC is reviewed as
general control strategy, with several implementations: Model
Algorithmic Control (MAC), Dynamics Matrix Control (DMC)
and Generalized Prediction Control (GPC). Other associated
topics are also reviewed: synthesis approach, optimization,
robustness, etc. A survey of MPC applications in industry can
be seen in [12].

680

Fig. 1. Model Predictive Control – use of model of system to predict
future behavior and optimally adjust inputs
A review of modern papers on Model Predictive Control
for HVAC systems is given in paper [13]. Authors make a
short introduction of HVAC systems and their modeling
(which is integral part of MPC). Then they discuss MPC and its
particulars (prediction horizon; inputs/outputs; constraints; cost
function; etc.). Finally, they give a review of 10 current
researches, comparing them on according to type (of model and
controller), type of cost function, solution method, optimality,
stage (computer simulation, experiment, actual building,
prototype) and special features.
Model Predictive Control and thermal modeling is used and
explained in paper [14], but it is not main a topic. Instead,
authors focus on establishing how perturbations of parameters
affect results and energy efficiency of HVAC system. They
build a thermal model of test building by Lumped Capacitance
Method (wall example shown in Fig. 2.; this is a very common
method for building thermal models for use in MPC) and build
parameter estimation errors into this model. They develop
Model Predictive Control system where they examine
influence of each perturbed parameter. This controller was
simulated over 2000 timesn performing Monte-Carlo
simulations to get result. Two parameters, thermal resistances
of floor and ceiling, are revealed to have majority of impact on
system results.
Model Predictive Control is extensively described in paper
[15]. Authors describe real-life application of MPC strategy for
control of HVAC system on University of California at
Merced, USA. They explain various topics, from modeling of
HVAC system, its elements and building, to control on several
levels of complexity. Many practical and expert details are
revealed, with overview of current research, which makes this
a very good starting point for researching Model Predictive
Control. Authors start with introduction to principles of MPC
and modeling of thermal processes in buildings. Control is
developed hierarchically, on two levels: high level, where they
model and control HVAC system, production of heating and
cooling water, distribution, etc.; and lower level which deals
with internal dynamic inside building and local elements of
HVAC system. Earlier version of this paper and research can
be found in [16].
Fig. 2. Lumped Capacitance Method – 3R2C substitute for wall
There are many papers dealing with Model Predictive
Control. Paper [17] describes a typical MPC problem and
explains various problems and challenges when developing a
control solution for HVAC system based on MPC. This
implementation uses Sequential Quadratic Programming (SQP)
to solve the optimization problem. Controller is based on
detailed and simplified thermal model of building. Authors
report that this solution is applicable in real-time control of
HVAC system. Paper [18] follows a typical structure of MPC
article, where in first part authors describe thermal model of
building (HVAC system, AHU – Air Handing Units and VAV
– Variable Air Volume units, room model based on variation of
Lumped Capacitance Method, etc.) and in second part they
deal with implementation of MPC (constraints, SQP method
for solving, etc.). Paper explains how Decentralized MPC
(DMPC) can be used for better results and how dual
decomposition method can be used to decentralize the system.
Authors of paper [19] follow a similar pattern but focus on
Stochastic MPC, where uncertainties of weather predictions are
built into model. Four control strategies are compared: Rule
Based Control (If-Then rules for control); Certainty
Equivalence MPC (uncertainty is neglected); Stochastic MPC
(where uncertainties follow Gaussian distribution); and
Performance Bound (optimal control with perfect knowledge
of future weather and long prediction horizon). Paper [20]
gives very detailed description of creating a Model Predictive
Controller for test plant, where MPC is combined with
feedback linearization (for linearization of input-output
behavior of cooling coil). Implementation of MPC for family
house is presented in [21]. Authors explain how to build a zone
model with Lumped Capacitance Method and then use it for
MPC. In paper [22] by same authors, focus is on indoor air
quality by ASHRAE (American Society of Heating,
Refrigerating and Air-conditioning Engineers) standards. In
[23] and [24] authors use subspace identification algorithm to
find thermal model of building from data. They test both their
identification strategy and control strategy on real building,
Czech Technical University. Authors stress impact of quality
of thermal model on prediction-based control strategies and
address question of feasibility of MPC in particular cases.
Paper [25] also deals with feasibility of Model Predictive
Control strategy. Authors question when is feasible to invest in
Model Predictive Controller for HVAC system on existing
building. To answer this, they compare energy balance of
building calculated by three methods: by use of some of
standard methods for static modeling (energy audit; energy
label or certificate); actual energy balance (energy loses; data
from Building Management System); and theoretical results
achieved by Model Predictive Control. There are two steps for
decision: first, if actual energy consumption is larger than
681
standard consumption (static model); second, if theoretical
MPC-based consumption is less than actual. Authors explain a
simplified modeling method that can be used in MPC strategy.
Another paper dealing with distributed MPC is [26].
Authors first give an explanation of Model Predictive Control
and develop a controller for a single room (or zone). They
focus much on possibility that MPC takes occupancy of zones
into consideration. Classic controllers (PID) usually use a
lower set-point for unoccupied periods (e.g. night-setup) which
saves energy; but MPC (especially if occupancy is known in
advance by schedule) can give much better results by ignoring
periods of inoccupation and preparing for future periods of
occupancy. In second part, authors develop their dMPC
(distributed MPC) that combines properties of centralized
control strategy (optimal results, because model of whole
building is taken into consideration) and decentralized control
strategy (lower computation costs and better robustness) by
making each of distributed controllers share data with its
neighbors. Paper [27] gives a review of distributed approach to
MPC. Authors have reviewed 35 approaches to dMPC (not
only in HVAC) and marked distinctive features, which enabled
them to speculate on future roadmap of dMPC research.
Usually MPC is used for control of HVAC system on high
level. In paper [28] MPC, in form of Inferential Model-based
Predictive Control Scheme (IMPCS), is used for control of a
single boiler, although again on high level, for defining set-
points, while PID control is used for direct control of actuators.
Detailed model of system is again needed, but in this case it is
much simpler model than in case of whole building. Model is
provided from hybrid method where structure of model
(including boiler and room) is based on first-principles
approach, while parameters are estimated by data-driven
approach. MPC is compared to classical ON/OFF control and
optimal control (Average Room Temperature-Based Control –
ARTBC), where results are very favorable to IMPCS. Authors
admit high costs of tuning this controller for each application.
A framework that couples building topology, estimation
and control routines is presented in paper [29]. Authors extend
their hybrid (gray-box) method for identification of thermal
model of building, which they developed in paper [30] (based
on Unscented Kalman Filter), and combine it with Energy
Management/Building Automation System. This enables
online identification through self-excitation that doesn't disrupt
normal routines of occupants (meaning that historical results of
control system are used to improve thermal model of building).
Finally, this online model is combined with Model Predictive
Control strategy. This would eventually lead to developing a
universal and automatic algorithm for identification and
control.
Paper [31] deals with variation of MPC called Economic
MPC (EMPC). In this case, the primary goal of control is not to
follow set-point exactly, but to use minimal amount of
resources. Here, price of electricity (or other energy sources) is
embedded directly into cost function. Drawback of EMPC are
high computation cost, which can be solved by reducing the
prediction horizon, which again results in poor results; and
occurrences of spikes (rapid changes of input signal), which
can be solved by imposing constraints on input, again affecting
cost of performance. To solve this drawback, authors
developed their method called Infinite-Horizon EMPC, which
they explain.
Authors of paper [32] deal with Generalized Prediction
Control (GPC), which is a variant of prediction control
developed during 80s. They propose three different model
types for thermal processes in building (LTV – Linear Time
Varying; BTV – Bilinear Time Varying; BTI – Bilinear Time
Invariant), all of which are inherently nonlinear. Two new
forms of GPC are presented: Modified incremental GPC
(MGPC) and Modified Non-Incremental GPC (MNIGPC); in
contrast to original (incremental) GPC and Non-Incremental
GPC (NIGPC). Also, it is possible to additionally adapt
(formulate) GPC to type of model: Linear Self Tuning Control
(LSTC) for LTV; Bilinear Self Tuning Control (BSTC); and
Fixed Bilinear Control (FBC). Detailed modeling of building
elements (room, wall, radiator, valve and boiler) is given and
results in bilinear product term that describes the whole system,
for which authors perform parameter estimation. Finally,
authors conduct simulations for each control algorithm (GPC,
NIGPC, MGPC, MNIGPC) and for each system representation
(LTV, BTV, BTI). Results favor incremental forms of GPC
and nonlinear forms of system.
Paper [33] describes a simple form of predictive control of
cooling in building, based on load-shifting. Three control
strategies will be compared: Night-Setup (NS), where cooling
is turned on only during period of occupancy; Load-Shifting
(LS), where building is pre-cooled during night (when energy
is cheaper) and cooled during day when interior temperature
rises above limit; and novel strategy called Demand-Limiting
(DL) that is proposed by authors. DL strategy also uses pre-
cooling during night, but instead of cooling during day, it uses
pre-cooling during day so that temperature never leaves limited
range. Goal of this method is to reduce peak loads, instead of
usual goal of lowering energy consumption. DL strategy is
based on model of building, which is explained in the paper.
In paper [34] three different control strategies are explained
and compared, applying on experimental HVAC system and
building. First, basic, strategy is classical On-Off control with
hysteresis (Bang-Bang), where controller just turns heater and
humidifier on and off when they leave defined ranges. Second,
advanced, strategy is based on On-Off controller, but considers
inertia of system, particularly temperature. Third, optimal
strategy is based on Model Predictive Control and achieves
best results, but expects large effort in modelling thermal
processes.
D. Fuzzy Control
Methods based on Fuzzy Logic Control are used quite often
in HVAC control, one of reasons being very simple
implementation of controller.
Very simple example of Fuzzy Logic in HVAC control is
shown in paper [35]. Authors use Fuzzy controller to turn on
and off HVAC units in four rooms of one building, in scenario
where there is not enough energy to keep them all turned on.
Fuzzy Controller must maintain the thermal comfort inside the
house, while avoiding situation where HVAC power load is
greater than available power (peak-load reduction). Authors
682
implement their controller in laboratory conditions. Similar
approach is used in [36]. Authors use two fuzzy controllers:
one analyses building and active energy consumers to provide
amount of available energy; and second distributes available
energy so that temperature in every zone stays in comfort level.
Paper [37] also deals with Fuzzy Logic. Authors first give
very extensive theoretical introduction to Fuzzy Logic
Controllers (FLC): Knowledge Base, Interface System,
Fuzzification Interface, Defuzzification Interface, which is
shown in Fig. 3. Ability to handle multi-criteria control
strategies is selected as one of most important properties of
FLC. Second part of paper deals with Genetic Tuning of FLC,
for which is also given an extensive description. Authors then
concentrate on their own contribution in form of Weighted
Multi-Criterion Steady-State Genetic Algorithm (WMC-
SSGA) for tuning of FLC. This strategy is then applied on two
different real-life test-sites, and one of them in different
scenarios: summer-season model and mid-season model.
Results of experiment are discussed with focus on energy
efficiency improvements and system stability.
Similarly to previous paper, [38] describes combination of
Fuzzy Logic Controller with Genetic Algorithms, which
authors call Evolutionary Fuzzy Rule-Based models. List of
rules is developed directly from data without previous model
information. Authors focus their work on keeping the
computational load (and complexity) at acceptable level, which
they realize by use of indices of rules. Additional benefit is that
with this method, physical interpretation of model is possible.
Proposed method is tested by modeling control of natural-gas
boiler.
Another combination of FLC and GA is given in paper
[39], where FLC is used to control valves that regulate water
and steam valves. Authors first give a detailed model of valves
and Air Handling Unit (AHU), and produce a multi-criteria
controller (temperature and humidity). This controller is also
able to re-tune itself to match new conditions, so they call their
controller Adaptive FLC (AFLC). They explain how GA is
used to modify Fuzzy Rule Matrix (FRM) that will provide
lowest possible RMS error. Results also show improvements in
response (rise times and settling times). Testing is implemented
in Matlab and Simulink.
A review of papers dealing with intelligent control
techniques in HVAC systems is presented in paper [40], where
by intelligent authors mean fuzzy-logic-based. First, a basic
introduction to fuzzy control is given, after which authors
continue with modified fuzzy techniques. They explain Neuro-
Fig. 3. Fuzzy Logic Controller and its elements
Fuzzy Systems (NFS) where Artificial Neural Networks
(ANN) are combined with FLC, and ANN is used to generate
or modify the Rule Base of FLC. Second approach instead of
ANN uses Genetic Algorithms (GA) for the same purposes,
and it is called Genetic Fuzzy Systems (GFS). In second part of
articles, a review of articles is given. Most of reviewed papers
deal with control of HVAC systems in various forms, but part
of them deal with improving of results of classical PID
controllers (by developing methods for auto-tuning). Another
review of fuzzy control strategies is given in paper [41], with
over 80 reviewed articles. Authors give an introduction to
Fuzzy Logic Control and continue to combination of FLC with
other control strategies: Genetic Algorithms; Clustering; Model
Predictive Control.
E. Other methods and general topic
Robust control is topic of paper [42]. Authors give only a
short introduction to robust control, explaining how in this type
of control uncertainties are assumed as part of system and
controller. Experimental HVAC system is built to test
performance, representing a typical HVAC system found in
residential buildings. Authors compare a typical control system
consisting of three separate SISO (Single Input Single Output)
PI (Proportional Integral) controllers with one MIMO
(Multiple Input Multiple Output) controller build on robust
theory principles. Results demonstrate improvement in form of
tracking times three times shorter that controlled by classical PI
control, applied on the same actuators.
A novel approach to HVAC control is given in [43] where
authors present a model-free control. Control is independent of
model or sensor information, and control actions are decided
upon votes of occupants. System is devised so that occupants
use their smartphones to vote about their current thermal
comfort (cold, neutral, hot), which triggers an action of HVAC
system (heating or cooling). There are two possible approaches
to voting: maximizing the number of comfortable people
(where few people are maximally comfort, but some people are
left very uncomfortable) or minimizing average thermal
discomfort (where nobody is perfectly comfortable, but nobody
is in discomfort). Additional energy save can be achieved by
enabling "drift", which means that indoor temperature is let to
drift to outdoor (inducing minimal thermal discomfort, but
largely reducing energy consumption).
Paper [44] describes a control system based on adaptive
feed-forward control strategy. Authors explain thermal
modeling of building by Lumped Capacitance Method
(resulting in graph mode of zones) and air-mass flow
equations. Goal of control is to move the equilibrium of system
closer to desired operating point, which is called Passivity
Based Control. Uncertainty of inputs is also taken into
consideration to achieve more robust control system. Method is
tested on simulation of four-room building, where model is
built by Computational Fluid Dynamics (CFD) software.
Paper [45] researches how to decentralize a control system.
Authors first elaborate centralized and decentralized control
systems – former can achieve optimal results, but are sensitive
to model correctness and problems with communication and
sensor failures; latter realize only suboptimal results, but can be
683
more robust. When dealing with HVAC systems in buildings,
decentralization is realized by splitting zones in such way that
several zones are controlled by one decentralized controller.
Usual approach to decentralizing is by heuristic methods, or
designer's "feel". Authors propose a systematic method based
on two factors: Coupling Loss Factor (CLF) which is
connected to degree of clustering (few clusters is better); and
Mean Cluster Size (MCS) which depends on average number
of elements in clusters (smaller clusters are better). Authors
explain how to divide zones in stages with use of combinatorial
methods, how to calculate CLF and MCS factors for each
stage, and how to select the optimal stage of clustering.
Paper [46] deals with general topic of comparison of
different control strategies. Authors propose that experimental
comparison can be misleading if conditions of experiment are
not the same: either different outside temperature conditions; or
not taking into consideration fundamental differences between
control strategies. Additionally, properties of measuring system
have to be taken into consideration, as separating influences of
other system in the building have to be, too. Paper focuses on
defining quantitative metrics for use of energy and for thermal
comfort of occupants. Resulting methodology is tested on two
different control strategies: classical control with Advanced
Scheduling and Model Predictive Control.
III. COMPARISON OF CONTROL STRATEGIES
All reviewed papers (excluding books and papers about
HVAC in general) can be seen in Table III. Papers are sorted
by order of appearance.

TABLE III. COMPARISON OF REVIEWED PAPERS FOR HVAC CONTROL

Classical Predict. Intelligent Model Identification Scope of
Other Improvement
Methods Methods Methods type approach control

Zones model
Data-driven

HVAC sys.
or element
Predictive

Nonlinear

principles

Complete
Reference

General

General
Genetic

Hybrid
On-Off

system
Linear
Topics

Other
Fuzzy

First-
MPC

ANN
Alg.
PID
Year

[5] 2012 x
[6] 2010 x x x x 40% improved tracking
[7] 2009 x x x x 82% improved efficiency
[8] 2009 x x x x
[9] 2008 x x x x
[10] 2005 x x x x
[12] 2003 x x
[13] 2013 x x
[14] 2011 x x x x
[15] 2012 x x x x 19% improved efficiency
[16] 2009 x x x x 24.5% improved efficiency
[17] 2011 x x x x
[18] 2011 x x x x 10.2% improved efficiency
[19] 2010 x x x 40% improved efficiency
[20] 2011 x x x x
[21] 2011 x x x x
[22] 2013 x x x x x
[23] 2010 x x x x 30% improved efficiency
[24] 2010 x x x x 10% improved efficiency
[25] 2011 x x x x x
[26] 2010 x x x x
[27] 2014 x
[28] 2010 x x x x 9-20% improved efficiency
[29] 2013 x x x x x 7.5% effic./50% track.
[31] 2012 x x x x 99.9% reduced comp. cost
[32] 2011 x x x x x
[33] 2007 x x x x 30% reduced peaks
[34] 2011 x x x x x x x 8.21% improved efficiency
[35] 2012 x x x
[36] 2009 x x x
[37] 2003 x x x x 30% improved efficiency
[38] 2000 x x x x
[39] 2013 x x x x
[40] 2010 x x x
[41] 2006 x
[42] 2008 x x x x x 300% improved efficiency
[43] 2013 x x 50% improved efficiency
[44] 2013 x x x x 16% improved efficiency
[45] 2011 x x x x x
[46] 2012 x x x x
40 1 6 3 21 7 4 1 8 3 20 11 12 10 9 14 9 11

684
 As explained before, regarding the control approach, papers
are divided in four categories (grey part of table): Classical
Control methods; Predictive Control Methods; Intelligent
Control Methods; and other methods and general topics.
Classical approach comprises control methods that have
been developed in past, e.g. On-Off and PID. These methods
are still used in majority of applications, but their popularity as
research subject is expectedly not strong anymore. Some
possible developments are in combining these methods with
variations of advanced approach.
Predictive methods present a major point of interest for new
papers, especially Model Predictive Control. MPC in HVAC
has been extensively researched in past (MPC as method exist
since 1970s), but development and new application are still
possible. Since efficiency of MPC depends on quality of
thermal model, a lot of study is put into developing better
models of buildings and HVAC systems. There are several
drawback of MPC in regard to its implement. First, MPC is
computationally intensive. Second, complex algorithm means
more room of errors. And third, for complex systems as large
commercial buildings, MPC has many parameters that have to
be tuned, so commissioning is hard. This means that there is
much effort to produce applications that can be used in
commercial products.
Intelligent control methods are also interesting for research,
especially those based on Fuzzy Logic Controllers, in
combination with Genetic Algorithms or Artificial Neural
Networks. They can give very good results and are relatively
easy to implements. But, judging from lack of implementation
in commercial products, there is still place for improvement
and these methods are still looking for their place on the
market.
Table III. also shows some information about type of model
that is used to represent the system, where such information is
stated in the paper. First category is model type, where models
are divided as linear or nonlinear. HVAC systems are usually
complex and nonlinear systems, but most papers either regard
them as linear or use some kind of linearization methods on
them.
Second division is based on type of approach that is used to
acquire the model. First-principles approach uses knowledge of
mathematics and physics to describe the systems dynamics,
thermal processes and properties of buildings. Such approach
enables useful insights in behavior of system, but it is usually
complex and requires specialized knowledge about buildings.
Data-driven approach uses numerical methods to produce the
model from data only. It can yield very good results in short
time, but it lacks interpretation of buildings interior behavior
and physical interpretation. Hybrid approach presents a
combination of previous two approaches, usually with use of
first-principles methods to provide a structure of model and
data-driven methods for parameter estimation.
Papers are then divided according to the scope or target of
control. Part of authors tries to target complete HVAC systems,
including the behavior of thermal conditions in building (e.g.
temperatures in each zone). Rest of authors tries to focus on
smaller scope: either only on HVAC system or its elements
(e.g. production of heat, control of boilers, etc.), where they
dismiss how produced energy will be distributed in building; or
on zones model (control of climate conditions inside the
building), presuming that necessary energy is available.
Lastly, Table III. gives some quantitative data about
improvements that resulted from implementation of these
control methods. These numbers must be taken with caution
and usually cannot be compared directly – context and
previous conditions of buildings and HVAC systems, and
scope of control must be regarded.
IV. CONCLUSION
As explained in introduction, as standard of living is
growing, so does the amount of energy used for heating,
ventilation and air conditioning. This signifies that methods for
efficient control of HVAC systems will only be more important
in future.
Classical approaches, especially PID controllers, are still
ubiquitous in most applications, whether domestic or
commercial. Advanced approaches, marked mostly by
predictive and intelligent control methods show very good
results when it comes to energy efficiency. Trivial problems of
advanced methods have been mostly solved. Next steps of
advanced control methods in HVAC should be finding ways to
make application commercially available and easier to
implement in different settings.
REFERENCES
[1] J. D. Kelso, “Building Energy Data Book 2011,” 2012.
[2] L. Perez-Lombard, J. Ortizb, and C. Pout, “A review on buildings
energy consumption information,” Energy Build., vol. 40, no. 3, pp.
394–398, Jan. 2008.
[3] V. Vakiloroaya, B. Samali, A. Fakhar, and K. Pishghadam, “A
review of different strategies for HVAC energy saving,” Energy
Convers. Manag., vol. 77, no. 1, pp. 738–754, Jan. 2014.
[4] R. McDowall, Fundamentals of HVAC Systems, 1st ed., vol. 40, no.
6. Elsevier, 2007.
[5] S. Kozak, “Advanced control engineering methods in modern
technological applications,” in Proceedings of the 13th International
Carpathian Control Conference (ICCC), 2012, pp. 392–397.
[6] N. Jain, R. J. Otten, and A. G. Alleyne, “Decoupled Feedforward
Control for an Air-Conditioning and Refrigeration System,” in
American Control Conference (ACC), 2010, pp. 5904–5909.
[7] M. Komareji, J. Stoustrup, H. Rasmussen, N. Bidstrup, P. Svendsen,
and F. Nielsen, “Simplified optimal control in HVAC systems,” in
2009 IEEE International Conference on Control Applications, 2009,
pp. 1033–1038.
[8] J. Liu, W. Cai, and G. Zhang, “Design and application of handheld
auto-tuning pid instrument used in HVAC,” in 2009 4th IEEE
Conference on Industrial Electronics and Applications, 2009, pp.
1695–1698.
[9] J. Wang, C. Zhang, and Y. Jing, “Application of an intelligent PID
control in heating ventilating and air-conditioning system,” in 2008
7th World Congress on Intelligent Control and Automation, 2008,
pp. 4371–4376.
[10] B. Tashtoush, M. Molhim, and M. Al-Rousan, “Dynamic model of
an HVAC system for control analysis,” Energy, vol. 30, no. 10, pp.
1729–1745, Jul. 2005.
[11] D. Bao-Cang, Modern Predictive Control. CRC Press, 2009.
685
[12] S. J. Qina, T. a. Badgwell, S. J. Qin, and T. a. Badgwell, “A survey
of industrial model predictive control technology,” Control Eng.
Pract., vol. 11, no. 7, pp. 733–764, Jul. 2003.
[13] R. Kwadzogah, M. Zhou, and S. Li, “Model predictive control for
HVAC systems — A review,” in 2013 IEEE International
Conference on Automation Science and Engineering (CASE), 2013,
pp. 442–447.
[14] S. Bengea, V. Adetola, K. Kang, M. J. Liba, D. Vrabie, R. Bitmead,
and S. Narayanan, “Parameter estimation of a building system
model and impact of estimation error on closed-loop performance,”
in IEEE Conference on Decision and Control and European
Control Conference, 2011, pp. 5137–5143.
[15] Y. Ma, A. Kelman, A. Daly, and F. Borrelli, “Predictive Control for
Energy Efficient Buildings with Thermal Storage,” IEEE Control
Syst., vol. 32, no. January, pp. 44–64, Feb. 2012.
[16] Y. Ma, F. Borrelli, B. Hencey, A. Packard, and S. Bortoff, “Model
Predictive Control of thermal energy storage in building cooling
systems,” in Proceedings of the 48h IEEE Conference on Decision
and Control (CDC) held jointly with 2009 28th Chinese Control
Conference, 2009, pp. 392–397.
[17] A. Kelman and F. Borrelli, “Bilinear Model Predictive Control of a
HVAC System Using Sequential Quadratic Programming,” in
International Federation of Automatic Control World Congress,
2011, pp. 9869–9874.
[18] Y. Ma, G. Anderson, and F. Borrelli, “A Distributed Predictive
Control Approach to Building Temperature Regulation,” in
American Control Conference, 2011, pp. 2089–2094.
[19] F. Oldewurtel, A. Parisio, C. N. Jones, M. Morari, D. Gyalistras, M.
Gwerder, V. Stauch, B. Lehmann, and K. Wirth, “Energy efficient
building climate control using Stochastic Model Predictive Control
and weather predictions,” in American Control Conference (ACC),
2010, 2010, pp. 5100–5105.
[20] J. Rehrl and M. Horn, “Temperature control for HVAC systems
based on exact linearization and model predictive control,” in 2011
IEEE International Conference on Control Applications (CCA),
2011, pp. 1119–1124.
[21] M. Vašak, A. Starčić, and A. Martinčević, “Model predictive
control of heating and cooling in a family house,” in Proceedings of
the 34th International Convention MIPRO, 2011, 2011, pp. 739–
743.
[22] M. Vašak and A. Starčić, “Model Predictive Control of a Heating,
Ventilation and Air Conditioning System,” in 36th International
Convention on Information & Communication Technology
Electronics & Microelectronics (MIPRO), 2013, pp. 913–918.
[23] J. Cigler and S. Privara, “Subspace identification and model
predictive control for buildings,” in 2010 11th International
Conference on Control Automation Robotics & Vision, 2010, no.
December, pp. 750–755.
[24] L. Ferkl, J. Siroky, S. Privara, and J. Sirok, “Model predictive
control of buildings: The efficient way of heating,” in 2010 IEEE
International Conference on Control Applications, 2010, pp. 1922–
1926.
[25] L. Ferkl, C. Verhelst, L. Helsen, A. Ciller, D. Komarkova, and D.
Kom, “Energy savings potential of a model-based controller for
heating: A feasibility study,” in 2011 IEEE International
Conference on Control Applications (CCA), 2011, pp. 871–876.
[26] P.-D. Morosan, R. Bourdais, D. Dumur, J. Buisson, and P. Moros,
“Distributed model predictive control for building temperature
regulation,” in American Control Conference (ACC), 2010, 2010,
pp. 3174–3179.
[27] R. R. Negenborn and J. M. Maestre, “Distributed Model Predictive
Control: An Overview and Roadmap of Future Research
Opportunities,” IEEE Control Syst., vol. 34, no. August, pp. 87–97,
2014.
[28] Z. Liao and A. L. Dexter, “An Inferential Model-Based Predictive
Control Scheme for Optimizing the Operation of Boilers in Building
Space-Heating Systems,” IEEE Trans. Control Syst. Technol., vol.
18, no. 5, pp. 1092–1102, Sep. 2010.
[29] P. Radecki and B. Hencey, “Online thermal estimation, control, and
self-excitation of buildings,” in 52nd IEEE Conference on Decision
and Control, 2013, pp. 4802–4807.
[30] P. Radecki and B. Hencey, “Online Building Thermal Parameter
Estimation via Unscented Kalman Filtering,” in American Control
Conference (ACC), 2012, 2012, pp. 3056–3062.
[31] D. I. Mendoza-Serrano and D. J. Chmielewski, “HVAC control
using infinite-horizon economic MPC,” in Proceedings of the IEEE
Conference on Decision and Control, 2012, no. 3, pp. 6963–6968.
[32] P. Kret, K. J. Burnham, T. Larkowski, and L. Koszalka, “Control
Optimisation of a Domestic Heating System,” in 2011 21st
International Conference on Systems Engineering, 2011, pp. 55–58.
[33] K. Lee and J. E. Braun, “Reducing Peak Cooling Loads through
Model-Based Control of Zone Temperature Setpoints,” in 2007
American Control Conference, 2007, pp. 5070–5075.
[34] K. Chinnakani, A. Krishnamurthy, J. Moyne, A. Arbor, and F. Gu,
“Comparison of energy consumption in HVAC systems using
simple ON-OFF, intelligent ON-OFF and optimal controllers,” in
2011 IEEE Power and Energy Society General Meeting, 2011, pp.
1–6.
[35] A. R. Al-Ali, N. A. Tubaiz, A. Al-Radaideh, J. A. Al-Dmour, and L.
Murugan, “Smart grid controller for optimizing HVAC energy
consumption,” in 2012 International Conference on Computer
Systems and Industrial Informatics, 2012, pp. 1–4.
[36] J. R. Villar, E. D. La Cal, and J. Sedano, “A fuzzy logic based
efficient energy saving approach for domestic heating systems,”
Integr. Comput. Aided. Eng., vol. 16, no. 2, pp. 151–163, Jul. 2009.
[37] R. Alcalá, J. M. Benítez, J. Casillas, O. Cordón, and R. Pérez,
“Fuzzy Control of HVAC Systems Optimized by Genetic
Algorithms,” Appl. Intell., vol. 18, no. 2, pp. 155–177, 2003.
[38] P. P. Angelov, R. a. Buswell, V. I. Hanby, and J. a. Wright, “A
methodology for modeling HVAC components using evolving
fuzzy rules,” in 2000 26th Annual Conference of the IEEE
Industrial Electronics Society. IECON 2000. 2000 IEEE
International Conference on Industrial Electronics, Control and
Instrumentation. 21st Century Technologies and Industrial
Opportunities (Cat. No.00CH37141), 2000, vol. 1, pp. 247–252.
[39] M. W. Khan, M. A. Choudhry, and M. Zeeshan, “An efficient
design of genetic algorithm based Adaptive Fuzzy Logic Controller
for multivariable control of HVAC systems,” in 2013 5th Computer
Science and Electronic Engineering Conference (CEEC), 2013, pp.
1–6.
[40] H. Mirinejad, K. C. Welch, and L. Spicer, “A Review of Intelligent
Control Techniques in HVAC Systems,” IEEE Int. Netw.
Infrastruct. Digit. Content, pp. 1–5, 2010.
[41] J. Singh, N. Singh, and J. K. Sharma, “Fuzzy modeling and control
of HVAC systems – A review,” J. Sci. Ind. Res., vol. 65, no. June,
pp. 470–476, 2006.
[42] M. Anderson, M. Buehner, P. Young, D. Hittle, C. Anderson, and
D. Hodgson, “MIMO Robust Control for HVAC Systems,” IEEE
Trans. Control Syst. Technol., vol. 16, no. 3, pp. 475–483, May
2008.
[43] S. Purdon, B. Kusy, R. Jurdak, and G. Challen, “Model-free HVAC
control using occupant feedback,” in 38th Annual IEEE Conference
on Local Computer Networks - Workshops, 2013, pp. 84–92.
[44] J. T. Wen, S. Mishra, S. Mukherjee, N. Tantisujjatham, and M.
Minakais, “Building temperature control with adaptive
feedforward,” in 52nd IEEE Conference on Decision and Control,
2013, vol. 1, no. 5, pp. 4827–4832.
[45] V. Chandan and A. G. Alleyne, “Optimal Control Architecture
Selection for Thermal Control of Buildings,” in American Control
Conference, 2011, pp. 2071–2076.
[46] A. Aswani, N. Master, J. Taneja, A. Krioukov, D. Culler, and C.
Tomlin, “Quantitative Methods for Comparing Different HVAC
Control Schemes,” in Proceedings of the 6th International
Conference on Performance Evaluation Methodologies and Tools,
2012, pp. 326–332.
686